tidygate: Interactively Gate Points

parse_gate_list = function(.data, my_df){
  
  # Comply with CRAN NOTES
  point = NULL
  
  .data %>%
    imap( ~ tibble(gate = .y, point = .x)) %>%
    reduce(.f = full_join, by = "point") %>%
    
    # Add all points
    full_join(tibble(point = as.character(1:nrow(my_df))), by = "point") %>%
    arrange(as.numeric(point)) %>%
    
    # Unite in case of a point belonging to multiple gates
    tidyr::unite(contains("gate"),
                 col = "gate",
                 sep = ",",
                 na.rm = TRUE) %>%
    
    # Replace NAs
    mutate(gate := if_else(gate == "", "0", gate)) %>%
    
    # Pull
    pull(gate)
}

#' @importFrom rlang quo_is_symbol
#' @importFrom grDevices colorRampPalette
#' @importFrom RColorBrewer brewer.pal
#' @importFrom viridis viridis
#' @import dplyr
#' @import tidyr
#' @importFrom scales rescale
#' @importFrom rlang :=
#' @importFrom graphics axis
#' @importFrom graphics legend
#' @importFrom graphics par
#' @importFrom utils head
#' @importFrom stringr str_pad
#' @importFrom scales alpha

pretty_plot_chr_int = function(.data,
                               .dim1,
                               .dim2,
                               .color = NULL,
                               .shape = NULL,
                               .size = NULL,
                               opacity = 1,
                               is_size_fixed) {
  # Comply with CRAN NOTES
  . = NULL
  color_hexadecimal = NULL
  
  # Get column names
  .dim1 = enquo(.dim1)
  .dim2 = enquo(.dim2)
  .color = enquo(.color)
  .shape = enquo(.shape)
  .size = enquo(.size)
  my_size_range = c(1, 3)
  
  
  
  .data_formatted =
    .data %>%
    
    # Define COLOR
    when(
      
      # If not defined
      pull(., !!.color) %>% unique %>% is.na() %>% all() ~ (.) %>% mutate(.color = "grey25", color_hexadecimal = "#3B3B3B"),
      
      # If continuous
      quo_is_symbol(.color) &&
        (.) %>%
        select(!!.color) %>%
        sapply(class) %in% c("numeric", "integer", "double") ~ {
          order_ = findInterval(pull(., !!.color), sort(pull(., !!.color)))
          (.) %>% mutate(color_hexadecimal = grDevices::colorRampPalette(viridis(n = 5))(n())[order_])
        },
      
      # If discrete
      quo_is_symbol(.color) ~ {
        how_many_colors = .data %>% distinct(!!.color) %>% nrow
        (.) %>%
          mutate(color_hexadecimal =
                   grDevices::colorRampPalette(RColorBrewer::brewer.pal(min(
                     9, how_many_colors
                   ), "Set1"))(how_many_colors)[factor(!!.color)])
      }
    ) %>%
    
    # Define SIZE
    when(
      # If not defined
      pull(., !!.size) %>% unique %>% is.na() %>% all() ~ (.)  %>% mutate(.size = 2),
      
      # If it is a number and not a column name
      is_size_fixed ~ (.)  %>% mutate(.size := !!.size),
      
      # If continuous
      quo_is_symbol(enquo(.size)) &&
        (.) %>%
        select(!!enquo(.size)) %>%
        sapply(class) %in% c("numeric", "integer", "double") ~ 	(.) %>%	mutate(.size := !!enquo(.size) %>% rescale(to = my_size_range)),
      
      # If discrete
      quo_is_symbol(enquo(.size)) ~ {
        warning("tidygate says: .size has to be a continuous variable. .size has been ignored")
        (.) %>% mutate(.size = 2)
      },
      
      ~ stop(
        "tidygate says: the parameter .size must be NULL, numeric or a symbolic column name"
      )
    ) %>%
    
    # Define SHAPE
    when(
      
      # If not defined
      pull(., !!.shape) %>% unique %>% is.na() %>% all() ~ (.)  %>% mutate(.shape = 19),
      
      # If continuous
      quo_is_symbol(.shape) &
        (.) %>%
        select(!!.shape) %>%
        sapply(class) %in% c("numeric", "integer", "double") ~ {
          warning("tidygate says: .shape has to be a discrete variable. .shape has been ignored")
          (.) %>% mutate(.shape = 19)
        }	,
      
      # If discrete
      quo_is_symbol(.shape) ~ (.) %>%	mutate(.shape := c(19, 17, 15, 18, 3, 4, 8, 10, 5)[factor(!!.shape)])
      
       
    )
  
  # Plot
  .data_formatted %>%
    {
      plot(
        (.) %>% pull(!!.dim1),
        (.) %>% pull(!!.dim2),
        xlim = range((.) %>% pull(!!.dim1)),
        ylim = range((.) %>% pull(!!.dim2)),
        bty = 'l',
        pch = (.) %>% pull(.shape),
        cex = (.) %>% pull(.size),
        col = (.) %>% pull(color_hexadecimal) %>% alpha(opacity),
        xlab = quo_names(.dim1) %>% paste(collapse = " "),
        ylab = quo_names(.dim2) %>% paste(collapse = " "),
        xaxt = 'n',
        yaxt = 'n'
      )
    }
  
  axis(1, tck = 1,  col.ticks = "light gray")
  axis(1,
       tck = -0.015,
       col.ticks = "black",
       labels = FALSE)
  axis(
    2,
    tck = 1,
    col.ticks = "light gray",
    lwd.ticks = "1",
    las = 1
  )
  axis(
    2,
    tck = -0.015,
    col.ticks = "black",
    las = 1,
    labels = FALSE
  )
  
  # Max length of the legends titles
  color_title = quo_name(.color)
  shape_title = quo_name(.shape)
  size_title = quo_name(.size)
  max_length_titles = max(nchar(color_title), nchar(shape_title), nchar(size_title))
  color_title = stringr::str_pad(color_title, width = max_length_titles, side = "both")
  shape_title = stringr::str_pad(shape_title, width = max_length_titles, side = "both")
  size_title = stringr::str_pad(size_title, width = max_length_titles, side = "both")
  
  # Add legend to top right, outside plot region
  inset_y = 0
  if (pull(.data, !!.color) %>% unique %>% is.na() %>% not() %>% all()) {
    legend(
      "topleft",
      inset = c(1.05, inset_y),
      legend = .data_formatted %>% 
        
        # If continuous
        when(
          pull(., !!.color) %>% class %in% c("numeric", "integer", "double") ~ arrange(., !!.color) %>% slice(1, n()) %>% pull(!!.color) %>% round(digits = 4),
          ~ arrange(., !!.color) %>% distinct(!!.color) %>% pull(!!.color) 
        ),
      pch = 19,
      col =  .data_formatted %>% 
        
        # If continuous
        when(
          pull(., !!.color) %>% class %in% c("numeric", "integer", "double") ~ arrange(., !!.color) %>% slice(1, n()) %>% pull(color_hexadecimal),
          ~ arrange(., !!.color) %>% distinct(color_hexadecimal) %>% pull(color_hexadecimal)
        ),
      title = color_title,
      box.col = "white",
      xjust = 0
    )
    inset_y = inset_y + distinct(.data_formatted,!!.color, .color) %>% nrow %>% magrittr::multiply_by(.1)
  }
  if (pull(.data, !!.size) %>% unique %>% is.na() %>% not() %>% all() &&
      (.data %>% select(!!enquo(.size)) %>% sapply(class) %in% c("numeric", "integer", "double"))) {
    legend(
      "topleft",
      inset = c(1.05, inset_y),
      legend = distinct(.data_formatted,!!enquo(.size)) %>% pull(!!enquo(.size)) %>% range,
      pch = 19,
      col = "black",
      pt.cex = my_size_range,
      title = size_title,
      box.col = "white",
      xjust = 0
    )
    inset_y = inset_y + 0.3
  }
  if (pull(.data, !!.shape) %>% unique %>% is.na() %>% not() %>% all()) {
    legend(
      "topleft",
      inset = c(1.05, inset_y),
      legend = distinct(.data,!!.shape) %>% pull(!!.shape),
      pch = distinct(.data_formatted,!!.shape, .shape) %>% pull(.shape),
      col = "black",
      title = shape_title,
      box.col = "white",
      yjust = 0
    )
    inset_y = inset_y + distinct(.data_formatted,!!.shape, .shape) %>% nrow %>% magrittr::multiply_by(.1)
  }
  
  
  
}

#' Get points within a user drawn gate
#'
#' @keywords internal
#'
#' @import dplyr
#' @import tidyr
#' @import tibble
#' @importFrom graphics plot
#' @importFrom purrr imap
#' @importFrom purrr map
#' @importFrom purrr reduce
#' @importFrom rlang quo_is_symbol
#'
#' @param .data A tibble
#' @param .dim1 A column symbol. The x dimension
#' @param .dim2 A column symbol. The y dimension
#' @param .color A column symbol. Color of points
#' @param .shape A column symbol. Shape of points
#' @param .size A column symbol. Size of points
#' @param opacity A number between 0 and 1. The opacity level of the data points
#' @param how_many_gates An integer. The number of gates to label
#' @param ... Further parameters passed to the function gatepoints::fhs
#'
#' @return A tibble with additional columns
#'
gate_interactive_chr_int <-
  function(.data,
           .dim1,
           .dim2,
           .color = NA,
           .shape = NULL,
           .size = NULL,
           opacity = 1,
           how_many_gates = 1,
           is_size_fixed,
           ...) {
    # Comply with CRAN NOTES
    . = NULL
    value = NULL
    
    # Get column names
    .dim1 = enquo(.dim1)
    .dim2 = enquo(.dim2)
    .color = enquo(.color)
    .shape = enquo(.shape)
    .size = enquo(.size)
    
    name = "gate"
    
    # my df
    my_df =
      .data %>%
      
      # Check if dimensions are NA
      check_dimensions(!!.dim1,!!.dim2)
    
    my_matrix	=
      my_df %>%
      select(!!.dim1,!!.dim2) %>%
      .as_matrix()
    
    # Add extra space to right of plot area; change clipping to figure
    if (pull(.data, !!.color) %>% unique %>% is.na() %>% not() %>% all() |
        pull(.data, !!.shape) %>% unique %>% is.na() %>% not() %>% all()|
        (pull(.data, !!.size) %>% unique %>% is.na() %>% not() %>% all() &&
         (
           .data %>% select(!!.size) %>% sapply(class) %in% c("numeric", "integer", "double")
         ))) {
      # Reset par on exit
      opar <- par(no.readonly = TRUE)
      on.exit(par(opar))
      
      # Set the new par
      par(mar = c(5.1, 4.1, 4.1, 8.1),
          xpd = TRUE,
          
          # Reduce tick length
          tck = -.01 
      )
    }
    
    
    # Plot
    my_df %>% pretty_plot_chr_int(
      !!.dim1,!!.dim2,
      .color = !!.color,
      .shape = !!.shape,
      
      # size can be number or column
      .size = !!.size,
      
      opacity = opacity,
      is_size_fixed = is_size_fixed
    )
    
    # Loop over gates # Variable needed for recalling the attributes later
    gate_list = map(1:how_many_gates,
                    ~ my_matrix %>% fhs(mark = TRUE, ...))
    
    # Save gate list
    temp_file = sprintf("%s.rds", tempfile())
    message(sprintf("tidygate says: the gates have been saved in %s", temp_file))
    gate_list %>% 
      map(~ attr(.x, "gate")) %>%
      saveRDS(temp_file)
    
    # Return
    gate_list %>% parse_gate_list(my_df)
    
  }

#' Get points within a user drawn gate
#'
#' @keywords internal
#'
#' @import dplyr
#' @import tidyr
#' @import tibble
#' @importFrom graphics plot
#' @importFrom purrr imap
#' @importFrom purrr map
#'
#' @param .data A tibble
#' @param .dim1 A column symbol. The x dimension
#' @param .dim2 A column symbol. The y dimension
#' @param gate_list A list of gates. Each element of the list is a data frame with x and y columns. Each row is a coordinate. The order matter.
#' @param ... Further parameters passed to the function gatepoints::fhs
#'
#' @return A tibble with additional columns
#'
gate_programmatic_chr_int <-
  function(.data,
           .dim1,
           .dim2,
           gate_list,
           ...
           ) {
    # Comply with CRAN NOTES
    . = NULL
    value = NULL
    
    # Get column names
    .dim1 = enquo(.dim1)
    .dim2 = enquo(.dim2)
    name = "gate"
    
    # my df
    my_df =
      .data %>%
      
      # Check if dimensions are NA
      check_dimensions(!!.dim1,!!.dim2)
    
    my_matrix	=
      my_df %>%
      select(!!.dim1,!!.dim2) %>%
      .as_matrix()
    
    # Loop over gates # Variable needed for recalling the attributes later
    gate_list_result = map(gate_list,
                           ~ .x %>%
                             when("data.frame" %in% class(.) ~ .as_matrix(.), ~ (.)) %>%
                             applyGate(my_matrix, .) %>%
                             which() %>%
                             as.character() %>%
                             
                             # Avoid error for empty gates
                             when(!is.null(.) ~ (.) %>% add_attr(.x, "gate")))
    
    # Return
    gate_list_result %>%
      
      parse_gate_list(my_df)
  
    
  }


.gate_chr_int = 		function(.dim1,
                           .dim2,
                           .color = NULL,
                           .shape = NULL,
                           .size = NULL,
                           opacity = 1,
                           how_many_gates = 1,
                           .group_by = NULL,
                           gate_list = NULL,
                           output_type = "chr",
                           ...)
{
  
  # Comply with CRAN NOTES
  . = NULL
  point = NULL
  
  # If gouping is complex multicolumn
  .group_by =
    .group_by %>%
    when(
      !is.null(.group_by) ~ matrix(ncol = length(.group_by) / length(.dim1)) %>%
        as.data.frame() %>%
        unite(".group_by", everything(), sep = "___") %>%
        pull(.group_by),
      ~ (.)
    )
  
    
  # Create tibble
  input_df = 
    tibble(
    .dim1 = .dim1,
    .dim2 = .dim2
  ) |> 
    mutate(.color = NA, .shape = NA, .size = NA)
  
  if(!is.null(.color)) input_df = input_df |> mutate(., .color = !!.color)
  if(!is.null(.shape)) input_df = input_df |> mutate(., .shape = !!.shape)
  if(!is.null(.size)) input_df = input_df |> mutate(., .size = !!.size)
  if(!is.null(.group_by)) input_df = input_df |> mutate(., .group_by = !!.group_by)
  
  
  unique_df = 
    input_df  %>%
    
    # Nesting is the case
    when(!is.null(.group_by) ~ nest(., data___ = .group_by),
         (.)) %>%
    
    distinct()
  
  # Interactive
  if(is.null(gate_list))
  result_vector = 
    unique_df |> 
    gate_interactive_chr_int(
      .dim1 = .dim1,
      .dim2 = .dim2,
      .color = .color,
      .shape = .shape,
      
      # size can be number of column
      .size =  .size,
      
      opacity = opacity,
      how_many_gates = how_many_gates,
      is_size_fixed = 
        class(.size) %in%  c("numeric", "integer", "double") &
        length(.size) < length(.dim1) &
        length(.size) == 1,
      ...
    )
  
  # Programmatic
  else if(is.list(gate_list))
    result_vector = 
    unique_df |> 
    gate_programmatic_chr_int(
      .dim1 = .dim1,
      .dim2 = .dim2,
      gate_list = gate_list
    )
  
  else 
    stop(
      "tidygate says: the parameter gate_list has to be NULL or a list of data frames"
    )
  
  # Convert
  if(output_type == "chr") result_vector = result_vector |> as.character()
  else if(output_type == "int") result_vector = result_vector |> as.integer()

  # Integrate maintaining order in case of nesting
  input_df %>%
    left_join(
      unique_df %>%
        mutate(gate = result_vector)
    ) |> 
    suppressMessages() |> 
    pull(gate)
}